[Another really great draft post from nikkoewan. If you missed his first one on the draft lottery, check that out here. - R]
The first post tried to look deeper into the mechanics of the draft lottery and game as an insight on WHICH positions "value" are best. The question of which position is most probably getting a top 3 pick is easy to answer - just look at the standings. But the question of which position gives us the best value for the position it is in, that's a different answer.
Now, the analysis showed that the 4 worst positions hold the most value. A slow decline in value is then observed for the next 4 positions (5th, 6th, 7th and 8th), followed by consecutive nosedives from 9 to 10, 10 to the positions 11,12,13 (13 is marginally better than 11,12) and lastly, the worst value is the 14th pick.
After identifying which positions hold the MOST value, we then look historically, which picks resulted in the best players.
A crucial element in this study is the idea of the lottery. The weighted lottery that uses 14 ping pong balls to correspond to the chances has been in place since 1994, so it's important that our data tackle only those until 1994 (since we are going to compare our results from the first study, to our results here). Of course, before the NBA consisted of 30 teams, there were just 13 lottery picks. I assumed that the 14th pick is no different than the 14th pick in a lottery because, as the analysis prior showed, being the 14th pick is no different than being 15th or 16th. Now, the question remains - how do we evaluate which players picked in the lottery performed the best?
We can't use raw box score statistics because as we all know, those are skewed either by pace(both by the season and by the team he played on). We can use PER, but I prefer not to. PER rewards inefficiency and undervalues defensive contributions. It accounts for production, but that single reason (rewards inefficiency and undervalues defensive contributions) makes it a very flawed statistic. Efficiency Differential is the best (in my opinion) because it tries to capture things which people say "can't be captured by statistics". Dean Oliver's Efficiency tried to take into account teamwork and synergy, which is exactly what we want. But that doesn't take into account playing time. Players have played few minutes and posted efficient numbers due to a small sample size. As such, I've arrived at using Total Win Share and WS/48, which is also a flawed statistic, which doesn't take into account synergy, but instead uses linear regression (using historical data) to fit coefficients into box score statistics to give us a measure of which box scores contribute to wins. If I have the time and the resource, I will try to look into using historical Eff Differential as a measuring tool.
How do we say that a player performed? Of course we want him to contribute greatly to wins. Therefore looking at WS/48 is a great way of analyzing this. However, just like efficiency, this measure doesn't take into account the sample size. Luckily we have Total Win Share. But again, TWS can be skewed towards players that play a ton of minutes yet produce little or marginally to wins. This also doesn't take into account longevity. To balance this out, I come up with my own system which I call Weighted Win Share Score. WWSS is simply the product between WS/48 and TWS/(total number of years the player SHOULD have played). Of course, we should note that WS/48 = TWS/(Total Minutes Played/48), so simplifying the product of WS48 and TWS/(years player SHOULD be playing) actually becomes (TWS)^2/((TMP/48)*(years player should have been playing)). That new equation, WWSS, gives a somewhat balanced weight between production, ability to stay on the court and longevity. This way, we can actually see who produced the best the longest while considering his ability to stay on the court. This rewards players who produce big for a few years (such as Oden), but also mitigate inability to stay on the court (again, such as Oden) and longevity (again like Oden). In total, this favors production(TWS and WS/48) more than longevity and staying power because I believe longevity and staying powers are highly dependent on luck (team situation, injury luck, coach luck) but production doesn't. Production is heavily dependent on the skills of a player - if he's good, he's good, no matter the circumstances.
Also, take note that (years player should have played) is calculated simply by subtracting 2012 to their draft year. Very simple. For players beyond 1998, the variable becomes a constant 14 because playing beyond 14 for ANY basketball player (let alone a superstar) is a HUGE thing, so players like Kobe and Duncan are rewarded (by reducing the denominator). Also, WWSS will be heavily skewed in the middle class (the 2004~2000 draft) because its right around the time when there is enough information to give us an accurate picture of the draft, yet not too much information that the data becomes skewed (like players beyond 1998).
Also, for comparison sake, if we assume that a player plays a WS/48 of .100. Why .100? Well, if I understood WS/48 correctly, it's really comparable to Win%. i.e. if a team plays 5 players with WS/48 of .100 for a full 48 minutes, then there total WS/48 of .500 (a simple summation) results in a record of 41-41. Thus, if a player has a WS/48 of .100, it means at HIS position, he is neither contributing to wins nor contributing to loses. This puts a WHOLE 'nother perspective on players such as Lebron James. Lebron James has a WS/48 of .336. Thus, if you put two players who play at a pace of WS/48 of .100, and two players who contribute negatively to wins (WS/48 of 0) together with Lebron and play them for a full 48 minutes for 82 games, your Win% would be 44-38. (If there is an error, please correct me via the comments).
Back to the control data, career WS/48 of .100, plays 28 minutes/game (a 6th man/starter - which is a reasonable assumption supposedly for Lottery Picks), and plays 8 seasons (2004 - the point at which there is sufficient data, but not enough to skew it) TWS = .100 * 28*8 = 22.4. Then his WWSS should be 22.4/8 * .100 = .28. Thus, .28 is our control point (this is the point of average-ness i.e. he neither made an impact nor be a negative contributor. ) Also, a quasi star (probable All Star) is someone who plays .150 WS/48 and plays around 34 minutes. for a WWSS of .765. Finally, a superstar is someone with WS/48 of .200 and plays around 36 minutes for a WWSS of 1.44
Now to the bat cave! *BANANANANA*
Data and Analysis
The bottom row is an average of that year, the last column is the average of that year. Looking at the bottom row, it's clear that 2003 lottery was a lottery beyond any other. It was the best of the best, trailed only by 1999 (whose lottery was headed by solid starters for many years like Brand, Hamilton, Davis, Odom, Miller, Sczerbiak, Marion, Terry and Maggette.) and the 1998 draft (headlined by a superstar for a long time in Nowitzki and Pierce, a superstar for maybe a good 8 years in Carter). Notice that the legendary 1996 draft (which produced Kobe, Nash, Iverson etc.) is behind. This is because Nash is outside the lottery (this study is ONLY for the top 14 picks) and their lottery produced duds like the highly inefficient Antoine Walker, Potapenko, Fuller and Wright. This is a study on the lottery, not the draft as a whole, therefore we disregard picks beyond the 14th (see assumptions above).
That is just a side note, our main concern is the right most column. That is the average WWSS over the span of 16 years of each pick. Notice that the 1st pick, comes out a whopping 0.93. Thus a 1st pick almost assuredly becomes a quasi star. This is true because since 2009, the 1st overall pick has produced Griffin, Rose, Oden (a superstar had he been healthy), Bargnani, Bogut, Howard, Lebron, Yao, etc. etc.
However, the decline from that point is HUGE. From .93, the next is .65 (3rd overall), .64(5th overall), .61 (4th overall) .57(9th overall. Funny food for the thought, the 9th pick has produced Dirk, Amare, Marion,Noah, Iggy and McGrady to lift this pick HIGHER than the 2nd pick) and finally the 2nd pick (.52).
Afterwards, it's a hodge podge of just below averageness. Meaning most players picked between 6-8 and 10-14 are contributing negatively to wins. Although there are some exceptions such as Roy (2006), Miller (1998), and of course Kobe(1996), most of the players picked in that range were BAD.
Well, from the previous analysis we've seen 2 things:
1.) Our own pick is safe (we're currently slotted at the 2nd position but I fully expect Washington to finish with a worse record than us giving us the 3rd position). Theoretically, the difference in value (w/ regards to getting a top 3 pick) between the top 4 positions are marginal (for numerical reasons, 3rd position is the "best"). Thus, we have no problems there (our own pick is maximized).
2.) for the MIN pick, what we need to pray for is that they bottom out to the 9th position. Value-wise, that's the best place that pick could be because the drop-off in value is HUGE from 9th to 10th (you don't experience a huge increase compared to the 11th position PLUS that's in a low sample size so 10th and 11th picks are basically the same in value).
This study further gives insight to what we should hope for in the upcoming draft. For starters, that 3rd position is 2nd only to the 1st overall pick in terms of historical data. Why? Maybe because the 2nd pick is pressured too much to perform the same level as the 1st pick, while the 3rd pick becomes the new standard for success for the 2 other top 5 picks. I don't know. What we do know is that, historically, the 3rd pick produced players that were productive for longer periods of times, compared to the 2nd pick. This is no longer quantifiable (reasons why the 3rd pick is better). But looking at the numbers, it's actually negligible.
Solving backwards for WWSS to come up with a WS/48 and TWS, our data comes up with, if we make the minutes constant for both the 2nd picks and 3rd picks (34 minutes/game) and averaging the number of years over the data(i.e. 2009 picks get 3 years, 2008 get 4 years, etc.. and averaging those x years) we get approximately 10 years. the 3rd pick produces a WS/48 of .138 for a TWS over a span of 10 years of 47 (remember, this is assuming he plays healthy for 82 games, and plays 34 minutes each game for all of those 10 years). While the 2nd pick produces a WS/48 of .123 for a TWS over a span of 10 years of 42. That's a difference of 5 games over 10 years, or half a game each year. Not too big. Compare that to the 1st pick, which over 10 years produced 56 wins or 9 games more than the 3rd, 14 games more than the 2nd. That's one more game each year that the 1st pick brings you, on average (of course, you don't get to pick 1st every year LOL).
For the MIN pick, there's something about that 9th pick. Historically, it's been good. But I don't know if that's explainable. Maybe after the top 5 sure picks, the next 3 picks are those that consist of "potential and/or skill without issues" and then followed by the 9th pick which drafts "potential and skill with issues" such as
in 09(Hill over DeRozan? (Err. Nope), Joe Alexander vs DJ Agustin in 08? (Maybe) Brandan Wright vs Noah in 09? (Maybe) I dunno. The evidence doesn't support that assumption (Flynn pick(produced in college), Gordon pick (skill AND potential), Rudy Gay and Brandon Roy etc..). So i can't think of any premise to support this claim. What I do know is that, the 9th position is the best among the worst. (9-14). Is this possible? With the injury to Rubio (hopefully he comes back healthy and strong next year without any lingering ACL tear effects), there MIGHT be a VERY SMALL chance(with Rubio out). That's if POR, UTH, PHX, MIL and CLE jump over MIN. those last 2 teams are HIGHLY unlikely to jump over MIN even with Rubio out.
To conclude, the dream is - hope we get the number 1 pick (whichever position we are, whether its 1,2,3 or 4) and for MIN pick to slide to that 9th position (where magic suddenly happens).